ISNet: Integrate Image-Level and Semantic-Level Context for Semantic Segmentation.
Zhenchao JinBin LiuQi ChuNenghai YuPublished in: ICCV (2021)
Keyphrases
- semantic level
- semantic segmentation
- context dependent
- pixel level
- input image
- superpixels
- scene classification
- image features
- single image
- image representation
- image content
- image classification
- low level
- weakly supervised
- test images
- conditional random fields
- pascal voc
- segmentation method
- multiscale
- region of interest
- semantic similarity
- image regions
- image retrieval
- similarity measure
- object categories
- object recognition
- natural language
- high level
- partial occlusion
- bounding box
- information extraction